I am currently using an Intel Pentium G850 (Sandy Bridge) dual-core CPU on one of my computers and I will be upgrading it to an Intel i5 2500 (Sandy Bridge) quad-core CPU. Are there any considerations I should make regarding my BOINC program?
In an abundance of caution, so nothing gets screwed-up, my plans are as follows:
1. Disable new E@H tasks and let the current tasks run to completion.
2. Uninstall BOINC.
3. Switch CPUs
4. Re-install BOINC (w/ E@H).
Am I over-complicating this? Are any of these steps really necessary? I appreciate any help you can give me. Thanks!
Copyright © 2024 Einstein@Home. All rights reserved.
Question about upgrading CPU from 2 to 4-core.
)
No, reinstalling BOINC or stopping any tasks is absolutely unnecessary.
Shutdown PC, replace CPU, power on and that's it. Boinc will immediately start working additional tasks on new cores.
Replacing whole motherboard might be a bit problematic for some OS-es (like Wintendo), though, but not boinc.
Thanks, Khangollo. I was
)
Thanks, Khangollo. I was hoping it would be that easy and problem-free.
RE: Thanks, Khangollo. I
)
It certainly is that easy.
Unless you already have the CPU and therefore don't have to spend money purchasing it, do you realise you can get a much greater boost in crunching performance by leaving the current CPU and just installing a suitable GPU - like a nVidia GTX650 for example - provided your motherboard has a spare PCIe2 x16 slot? You would save yourself the task of removing and replacing the heat sink and the tricky bit of making sure you correctly applied new thermal paste. Much easier to insert a graphics card and install drivers :-).
A GTX650 is very miserly on power and costs little more than $100 so your current PSU may well be completely adequate for the extra load. The performance boost would be around a factor of 10 times. You are currently doing BRP4 tasks on the CPU and they are taking more than 10 hours. If you did them on a GTX650 GPU, you could expect to do 2 tasks in around 1 hour.
Cheers,
Gary.
Thanks Gary! You've given me
)
Thanks Gary! You've given me some very useful information which I'll be considering in the near future, too.
I just recently "built" the two computers I'm using for BOINC. Both use the same brand/model motherboard, memory, PSU, etc. I recently purchased the quad-core CPU, which I'll install as soon as it arrives. Unfortunately, I did not realize that a suitable GPU, such as you mentioned, would drastically increase processing efficiency on the BRP4 tasks. I erroneously assumed that the CPUs would be more efficient at any task.
So, that brings up another question: In the interests of improving my contribution efficiency to E@H , until I install a suitable GPUs on the computers, would my contribution be more useful to the project if I simply deselected BPR4 tasks in my computing preferences for now?
As for messing with thermal paste and heat sinks, I've been doing some experimentation with CPU cooling devices and air-flow anyway. I consider any chance to open-up the covers and play around as "fun." :-)
The i5 2500 will absolutely
)
The i5 2500 will absolutely run circles around the Pentium G850. Not only will you be doing more tasks at once, but the completion time for each will be reduced.
Having said that, it can't compete with even a modest graphics card in terms of WU processing.
If your only goal is E@H AND you can only do one upgrade (cpu or gpu), then the GTX650 as suggested is the way to go.
Ideally however you're better off all around with the i5 2500 and the GTX650.
RE: In the interests of
)
Maybe. I have a suitable/reliable GPU and I disabled the CPUs from running BRP4 tasks so they can concentrate on work that cannot be done by GPUs. Since GPUs do BRP4 tasks so much faster I figured this would increase the speed as a whole (in the small way of my computer's work).
One can argue that CPUs are still the most reliable computation device (as was explained elsewhere) and discoveries have been made by them in this good project, but our habit of doing every working unit twice should take care of the reliability problem, no?
So why, dear admins, do we not make this approach the norm? Or are there too few GPUs among project contributors to handle the data?
RE: So why, dear admins,
)
My view on this is that we should not deprive, by default, CPU-only users of the chance to make a pulsar discovery and get one of those cool discovery certificates :-). Seriously, I think many users would fail to notice they have to do something actively to participate in this search, so it's safer, IMHO, to keep the search enabled by default also for CPU-only users.
HBE